How to run Facebook ads for a brand new ecommerce store

Launching Facebook ads for a new store comes down to three numbers you should know before you spend anything: the ROAS that breaks you even, the daily budget the learning phase actually requires, and the number of creatives you can ship each week. Get those wrong and no amount of tinkering with targeting will save the account.

What changed, and why old launch guides mislead

In February 2025 Meta merged Advantage+ Shopping and manual sales campaigns into a single Sales flow. Advantage+ stopped being a campaign type you pick and became a state your campaign is in. Turn off budget, audience or placement automation and, per Jon Loomer's reading, any one of those changes turns Advantage+ off — so the campaign reverts to manual without telling you it has.

That matters for a launch because most published walkthroughs still describe choosing between two campaign types on a screen that no longer offers that choice. The Existing Customer Budget Cap, which used to let you reserve a slice of spend for people who had already bought, was removed outright rather than deprecated. If a guide tells you to set it to 10%, that guide is describing a button that's gone.

Control has continued to narrow since. Placement selection at ad set level is now largely automated, leaving account-level settings as the remaining lever. We have not put a date on that one: the change is visible in the interface, and Meta has not documented it anywhere we can cite. The practical reading is simple: the settings you can still change are fewer and matter less, and the things you bring to the auction — your offer, your margin and your creative — matter more.

How much do you actually need to spend to launch?

Meta optimises an ad set using the conversions it receives, and it needs roughly 50 optimisation events per ad set per rolling seven days to leave the learning phase. That figure's at ad set level, not ad level, and it counts pixel, Conversions API and modelled conversions together.

Nobody writes down what that implies in money, so here's the arithmetic. Your weekly budget floor per ad set is fifty times whatever you are willing to pay for a purchase.

A store aiming for a $30 cost per purchase needs about $1,500 a week per ad set, or roughly $214 a day, to give the ad set a fair chance of exiting learning. At $60 a purchase it is $3,000 a week. That's the arithmetic that decides whether a launch is viable, and it's the reason a $20-a-day test of four ad sets tells you nothing: every one of them stays in learning indefinitely, and learning-phase performance is not representative of anything.

What ROAS do you need before any of this is profitable?

Break-even ROAS is one division, and almost everyone runs it on the wrong number. The usual shortcut divides by gross margin, which leaves out shipping, pick-and-pack, payment fees and returns. Those are real costs that scale with orders, and including them typically moves the break-even point by more than a full turn of ROAS.

Work yours out below. The number you get is the floor a campaign has to clear before it contributes anything, and it is also your maximum cost per acquisition on a first order.

There is a longer version of this argument in the break-even ROAS piece, including what belongs in contribution margin and what does not.

How should a new account be structured?

Consolidated. The instinct to build five ad sets around five audiences comes from an era when targeting was the lever, and it now works against you: each ad set needs its own fifty conversions, so splitting a budget five ways multiplies the learning problem by five while dividing the signal.

1. 01

   One sales campaign

   Objective: sales, optimising for purchases. Not link clicks, not add-to-cart, unless you genuinely can't generate fifty purchases a week at any budget — in which case optimise for the closest upstream event and treat the result as directional.

2. 02

   One or two ad sets, not five

   Start with one broad ad set. Add a second only when you have a specific hypothesis that needs isolating, and only when you can fund both above the floor.

3. 03

   Six to ten creatives in the ad set

   Distinct concepts, not six crops of one image. Meta needs variation to have something to choose between, and you'll need it to learn which angle works.

4. 04

   Leave attribution on the default

   Seven-day click and one-day view. Change it and you'll lose comparability with every benchmark and every past period, for no gain during a launch.

5. 05

   Set the pixel and the Conversions API together

   Browser-side alone loses a meaningful share of events to tracking prevention, and those missing events come straight out of the fifty you need.

Should you use Advantage+ or manual campaigns?

Pick one and commit. The largest public body of evidence on this is a set of 640 incrementality experiments run by Haus across advertisers averaging $14M a year in Meta spend, and it says three useful things.

Read that last number carefully, because it's the one that changes what you do on launch day. Hedging measured worse than either commitment. The likely mechanism is that a split budget starves both sides of the conversions they need, which is the same learning-phase problem in a different coat.

For a brand new account with no conversion history, Advantage+ is the reasonable default: it has more room to explore, and you don't have audience insight worth encoding yet. Revisit the choice once you have a few hundred purchases, and revisit it with a holdout test rather than by comparing two dashboards.

How many creatives do you need in the first month?

More than feels reasonable. Across $1.29B of Meta spend and 578,750 creatives, Motion found that roughly 5% of creatives are winners — defined as taking at least ten times the account's median creative spend and at least $500. That threshold sits at the 92.3rd percentile.

The same dataset found enterprise advertisers — which Motion defines as spending upwards of $1M a month on Meta — shipping about 18.8 net-new creatives a week. You don't need that on day one, but you do need a rate you can sustain, because the launch doesn't end when the campaign goes live. Plan production before you plan budget: an account that can ship four concepts a week will find its first winner in about five weeks, and an account that ships one will take twenty.

On what to make: the fastest way to a first slate of concepts is to read what is already running in your category. Competitors' ad libraries are public, and the length of time an ad has been live is the closest thing to a performance signal you can see from outside. Reading a competitor ad library properly covers what the public record does and does not contain, and the angle taxonomy is a way to make sure your first ten are ten different arguments rather than ten different images.

What should you look at in week one, and what should you ignore?

Ignore almost everything. An ad set in learning produces numbers that don't predict its steady state, and the strongest temptation in a launch is to act on them.

Hook rate deserves a specific warning because it is the metric most likely to send a new advertiser after the wrong fix. It measures three-second plays over impressions, which tells you whether the first frame stopped the scroll. It's a genuinely useful diagnostic and it isn't a conversion predictor — high hook rate with weak sales usually means the audience or the landing page is wrong, not the hook. Hook rate and hold rate goes through both definitions, including the two competing denominators that make cross-tool comparisons meaningless.

A launch checklist

The part most launches get wrong

Almost every failed launch we have looked at failed on arithmetic rather than execution. The budget was below the learning floor, or the break-even ROAS was computed on gross margin and the real one was unreachable, or the creative pipeline produced three assets and stopped. None of those get fixed by better targeting, and all three are visible before a single dollar is spent.

Work out the three numbers first. If they do not clear, the honest answer is that the channel isn't ready for this product yet, and finding that out on a spreadsheet is considerably cheaper than finding it out on an invoice.